Understanding the Information Propagation Effects of Communication Topologies in LLM-based Multi-Agent Systems

TL;DR

This paper investigates how communication topologies in large language model multi-agent systems influence information flow, proposing a causal framework and a novel topology design to optimize performance, robustness, and communication costs.

Contribution

It introduces a causal analysis framework for topology effects and proposes EIB-leanrner, a new method balancing error suppression and information diffusion.

Findings

Moderately sparse topologies optimize task performance.

EIB-leanrner outperforms existing topologies in effectiveness.

Sparse topologies can effectively suppress error propagation.

Abstract

The communication topology in large language model-based multi-agent systems fundamentally governs inter-agent collaboration patterns, critically shaping both the efficiency and effectiveness of collective decision-making. While recent studies for communication topology automated design tend to construct sparse structures for efficiency, they often overlook why and when sparse and dense topologies help or hinder collaboration. In this paper, we present a causal framework to analyze how agent outputs, whether correct or erroneous, propagate under topologies with varying sparsity. Our empirical studies reveal that moderately sparse topologies, which effectively suppress error propagation while preserving beneficial information diffusion, typically achieve optimal task performance. Guided by this insight, we propose a novel topology design approach, EIB-leanrner, that balances error suppression and beneficial information propagation by fusing connectivity patterns from both dense and sparse graphs. Extensive experiments show the superior effectiveness, communication cost, and robustness of EIB-leanrner.